Almost twenty surveys on voting intentions have already been published since the beginning of September. This abundant production gives a strong rhythm to the media agenda. The measuring instrument has become an actor in the electoral campaign which puts pressure on all the candidates. A nervousness exacerbated by the irruption of the “Zemmour phenomenon” – which did not declare itself a candidate – and the publication of polls placing him on the podium of the first round, even qualifying him for the second. The passionate debates lead us to take a methodological step back from what “voting intentions” measure.
If the polling institutes define three or four questions relating to the intentions to vote (registration on the electoral rolls, certainty of going to vote then choice of first and second round …), there are differences in the evaluation of “certainty” go to vote: Harris Interactive measures it with a scale in four categories, IFOP and OpinionWay with a scale of 1 to 10, Ipsos-Sopra Steria and Elabe from 0 to 10. This can have strong consequences: the Harris Interactive survey of October 6 (the one with a Zemmour voting intention of 17% or even 18%) indicates a certainty rate of going to vote of 72% while that of Ipsos of the 1er October (Zemmour voting intention at 15%) indicates a 50% certainty of going to vote.
This is not just a simple methodological detail. Indeed, “voting intentions” relate (with some exceptions) only to people “certain” to vote in order to obtain a “chemically pure” measure of electoral choices. Except that the various polls do not put the cursor of “certainty” in the same place: for some (Ipsos-Sopra Steria and OpinionWay), it is only note 10 which defines the certainty of going to vote, for others (Elabe) it’s from the 8th.
This rather virtuous approach despite the methodological differences between institutes, reduces the sample of respondents to the question of “voting intentions”, sometimes by half. Thus the Harris Interactive survey which places Eric Zemmour in the second round is based on approximately 700 people while the representative sample was at the base of 1,310 French people, or 1,062 people registered on the electoral rolls. Hence the importance of having large samples.
Beyond these differences and these biases, it should be remembered that when the voting intentions are only asked of voters “certain” to vote, they are not the reflection of the potential electoral population and in particular not of the voters. undecided, uncertain, who often do not decide until late. If the pollsters make a significant methodological effort so that people “certain” to vote do not constitute a too deformed subset of the potential electoral population, this subset nevertheless presents a more sociological profile more socially favored, more senior in particular, and more politically motivated than the general population.
You have 51.48% of this article to read. The rest is for subscribers only.